Overview
My research interests are in Artificial Intelligence and Machine Learning.
In particular, I focus on combining Computer Vision and Natural Language Processing to advance the field of Spatial Intelligence - machines that understand what objects are, as well as their size, shape, and location.
I believe in a future of machines that understand the structure and content of the world around them, and use this to reason, plan and act in their environment.
Formally, my research areas include world modelling, self-supervised representation learning, and object-centric vision.
Research Projects
Applied Research
(Conducted as Founder & CEO of Spatial Intelligence and as Research Scientist at PHilita.)
- Spatial Context Protocol: AI agents that understand, navigate, and control physical buildings in real time.
- Compositional 3D Scene Building: Generating 3D scenes and objects from natural language descriptions.
- Spatio-temporal modelling: Predictive epidemiology and public transport optimisation during the pandemic. (PHilita x TfL, presented at the European Transport Conference, Milan, 2023).
Fundamental Research
(Conducted in collaboration with the University of Bristol, as Research Scientist at PHilita, and as Founder & CEO of Spatial Intelligence.)
- Spatial Intelligence: From structured representations to object-centric world models.
- JEPAs: Joint-Embedding Predictive Architectures and latent-variable energy based models for self-supervised representation learning and world modelling.
- Object Representation Learning: Semantic, spatial, and relational object understanding for object-centric world models.
- Latent Cycle Transformers: Variable-length continuous reasoning over heirarchical representations for true “thinking” models.
- Spatio-temporal modelling: MGCRNN - Multi-graph convolutional, recurrent neural networks for space-time series prediction.